RANDOM CONNECTIVENESS BETWEEN SENSORY BUMP ATTRACTORS AND MIXED SELECTIVE NEURONS UNDERLIES FEATURE BINDING: PRELIMINARY EVIDENCE

Abstract

We are constantly faced with a gazillion objects around us. And each of these objects has a myriad of features (e.g., color, size, shape, location). How does our brain combine features selectively to specific objects and represent each of them as a unified whole? The process underlying this deceptively simple task is called feature binding, and it is a key feature of the working memory. Despite theoretical advances, the underlying neural mechanism is still uncertain. This study aimed to experimentally investigate the neural mechanism of feature binding by empirically evaluating two recent models centered on alternative hypotheses – selective synchronization through bump attractors and mixed selectivity with random connectiveness. The data used for this study came from prefrontal recordings of an adult macaque which performed a delayed response task requiring the binding of color and location information. The result of this study provides preliminary evidence for the presence of bump attractor dynamics in color selective and mixed selective neurons. By looking at this evidence through the lens of past studies, this study hints at a hybrid model of WM where sensory bump attractors and mixed selective neurons work in tandem to produce feature binding through random connectiveness between them. This model, by providing a reconciliation of the two alternative views of feature binding, opens a new avenue of research on working memory.